Performance hit % for 5 disk RAIDZ2?

Status
Not open for further replies.

Chip Sprague

Dabbler
Joined
Aug 21, 2016
Messages
21
4TB Drives.

I'm fine with 12TB usable space. I'd rather be able to take a 2nd disk failure if it only means something like 10% to 15% performance hit on IO. Any way to quantify the approximate performance hit vs 4 or 6 disks?
 

Mirfster

Doesn't know what he's talking about
Joined
Oct 2, 2015
Messages
3,215
In a nutshell, regardless of the RaidZ level (1, 2, 3) you are talking about a single vDev; so it would be the IOPS performance of a single drive.

Any other differences are pretty negligible really, but general thoughts are RaidZ1 is a little faster that RaidZ2, but I don't know if I would say 15% faster...

However, plenty of other factors come into play (like file size, single large file or bunch of small files or sequential vs random reads) that are not easily added to a single calculation.

Have a read of this: "A Closer Look at ZFS, Vdevs and Performance"
 

joeschmuck

Old Man
Moderator
Joined
May 28, 2011
Messages
10,996
Any way to quantify the approximate performance hit vs 4 or 6 disks?
You could create a vdev in both configurations and run some testing. Otherwise I agree with @Mirfster above, RAIDZ1 is a little bit faster than RAIDZ2, and you are likely to see the difference on a 4 disk vdev more so than a 6 disk vdev. Now these would be speed differences measurable within FreeNAS, not over a network connection. So this would not be an issue for a home system.
 

Chip Sprague

Dabbler
Joined
Aug 21, 2016
Messages
21
Thanks for the feedback and that link was great. I get why 6 in a 3+3 setup would be faster. I don't see any reason 5 disks in a RAIDZ2 config w/two disk tolerance would be slower than 4 w/one disk tolerance. Am I missing something?
 

Ericloewe

Server Wrangler
Moderator
Joined
Feb 15, 2014
Messages
20,194
Thanks for the feedback and that link was great. I get why 6 in a 3+3 setup would be faster. I don't see any reason 5 disks in a RAIDZ2 config w/two disk tolerance would be slower than 4 w/one disk tolerance. Am I missing something?
RAIDZ2 is "slower" because it requires more complex calculations for the second set of parity.

The parity has to be structured so as to allow the original data to be reconstructed from any two available blocks, out of the four that exist.

This is not a problem for most users in realistic scenarios with modern hardware.
 
Status
Not open for further replies.
Top